-
Notifications
You must be signed in to change notification settings - Fork 0
/
demo.py
85 lines (61 loc) · 2.17 KB
/
demo.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
# coding=utf8
import logging
import warnings
import torch
from openprompt.data_utils import InputExample
from openprompt.plms import T5TokenizerWrapper
from openprompt.prompts import MixedTemplate
from transformers import RobertaTokenizer, T5ForConditionalGeneration, \
T5Config
from openprompt import PromptDataLoader, PromptForGeneration
warnings.filterwarnings("ignore")
logger = logging.getLogger()
logger.setLevel(logging.CRITICAL)
logger = logging.getLogger()
logger.setLevel(logging.ERROR)
model_config = T5Config.from_pretrained(r'codet5-base')
plm = T5ForConditionalGeneration.from_pretrained(r'codet5-base')
tokenizer = RobertaTokenizer.from_pretrained(r'codet5-base')
WrapperClass = T5TokenizerWrapper
promptTemplate = MixedTemplate(model=plm, tokenizer=tokenizer,
text='The problem description is: {"placeholder":"text_a"} The code snippet is: {"placeholder":"text_b"} {"soft":"Generate the question title:"} {"mask"} ',
)
model = PromptForGeneration(plm=plm, template=promptTemplate, freeze_plm=False,
tokenizer=tokenizer,
plm_eval_mode=False)
model.load_state_dict(torch.load('../model/pytorch_model.bin'))
example = []
desc = 'There is "bid" and "ask", but no actual stock price.'
code = """
import yfinance as yf
stock = yf.Ticker("ABEV3.SA")
data1= stock.info
print(data1)
"""
example.append(
InputExample(
guid=0,
text_a=desc,
text_b=code
)
)
data_loader = PromptDataLoader(
dataset=example,
tokenizer=tokenizer,
template=promptTemplate,
tokenizer_wrapper_class=WrapperClass,
max_seq_length=512,
decoder_max_length=64,
shuffle=False,
teacher_forcing=False,
predict_eos_token=True,
batch_size=1,
)
generated_texts = []
groundtruth_sentence = []
guids = []
for batch in data_loader:
with torch.no_grad():
_, output_sentence = model.generate(batch, num_beams=10,num_return_sequences=10)
generated_texts.extend(output_sentence)
print(generated_texts[0])